arXiv:2602.14612v4 Announce Type: replace-cross 
Abstract: Answering natural-language questions over multi-hour audio requires both event recognition and temporal grounding. Current large audio-language models perform well on short clips, but are limited by context length, query-time cost, and weak temporal localization. We present LA-RAG (Long Audio-Retrieval Augmented Generation), a structured framework that converts continuous audio into timestamped event records using an open-vocabulary Audio Grounding Model (AGM), stores them in a SQL event database, and answers queries through intent-aware retrieval followed by LLM-based generation. LA-RAG supports offline grounding mode, where long recordings are pre-indexed for low-latency QA, and inference-time grounding mode, where query-conditioned grounding is performed for shorter open-ended clips. We create 24-hour Home-IoT and Industrial-IoT audio benchmarks and augment CASTELLA, a real-world audio moment retrieval dataset with QA pairs. In offline grounding mode, LA-RAG achieves 76.88% overall accuracy on Home-IoT and 71.10% on Industrial-IoT, with average query latencies below 0.6 seconds. In inference-time grounding mode, state-of-the-art LALMs achieve competitive event-detection accuracy on CASTELLA-QA but low temporal detection F1. We further show that LALMs augmented with our structured retrieval metadata achieve consistent temporal detection improvements, with F1 gains of 11-17% across baseline models with improved latency. These results show that explicit timestamped grounding and structured retrieval provide a practical complement to generative audio-language models for deployment-oriented long-audio QA.

تم تقديم إطار عمل جديد يسمى LA-RAG (توليد استرجاع الصوت الطويل المعزز) لتحسين الإجابة على الأسئلة حول التسجيلات الصوتية الطويلة. يستخدم هذا النظام نموذج تحديد الصوت لتحويل الصوت المستمر إلى سجلات أحداث مؤرخة، مما يتيح استرجاعًا فعالًا وتوليد ردود على الاستفسارات بلغة طبيعية.

Se ha presentado un nuevo marco llamado LA-RAG (Long Audio-Retrieval Augmented Generation) para mejorar la respuesta a preguntas sobre grabaciones de audio largas. Este sistema utiliza un Modelo de Localización de Audio para convertir audio continuo en registros de eventos con marcas de tiempo, lo que permite una recuperación eficiente y generación de respuestas para consultas en lenguaje natural.

Un nouveau cadre appelé LA-RAG (Long Audio-Retrieval Augmented Generation) a été introduit pour améliorer la réponse aux questions sur de longs enregistrements audio. Ce système utilise un modèle de mise au sol audio pour convertir l'audio continu en enregistrements d'événements horodatés, permettant une récupération efficace et une génération de réponses pour des requêtes en langage naturel.

A new framework called LA-RAG (Long Audio-Retrieval Augmented Generation) has been introduced to enhance question answering over lengthy audio recordings. This system utilizes an Audio Grounding Model to convert continuous audio into timestamped event records, enabling efficient retrieval and response generation for natural-language queries.